With recent advancements in next-generation data storage, especially in biological molecule-based storage, insertion, deletion, and substitution (IDS) error-correcting codes have garnered increased attention. However, a universal method for designing tailored IDS-correcting codes across varying channel settings remains underexplored. We present an autoencoder-based approach, THEA-code, aimed at efficiently generating IDS-correcting codes for complex IDS channels. In the work, a disturbance-based discretization is proposed to discretize the features of the autoencoder, and a simulated differentiable IDS channel is developed as a differentiable alternative for IDS operations. These innovations facilitate the successful convergence of the autoencoder, producing channel-customized IDS-correcting codes that demonstrate commendable performance across complex IDS channels, particularly in realistic DNA-based storage channels.

Abstract In big data analysis, a large quantity of OLAP and aggregate queries exists, which have much stronger semantic context relationships (e.g. drill down and roll up) than generic SQL queries. Caching query results in memory playing an important role in accelerating data queries. Nevertheless, traditional query caching schemata neither fully utilize the features of OLAP, such as drill down and roll up semantics, nor compress the cached results, as the memory space is limited. In this paper, we propose a succinct, adaptive semantic OLAP caching, where the cache items are the cube lattice equivalence classes with only the bounds in a class stored. With further queries, the bound ranges are extended or expanded, indicating more query-answering ability which is assessed by the proposed covering capacity. The bounds of equivalence classes that more covering capacity are preferentially preserved in caching. We further empower our cache with some inference ability to derive more new data cells without posing extra queries and develop efficient query and update algorithms. The extensive experimental evaluation is conducted on synthetic and real data sets with various parameter settings. Our cache outperforms the common caching like LRU and LFU. Furthermore, it is robust to the non-repeated-pattern queries, still with a 30% hit ratio.

Deep neural networks have led to the development of various data-driven architectures for image compressive sensing (CS), generally categorized into black-box models and deep unfolding networks (DUNs). DUN approaches have garnered attention due to their interpretability. However, most existing methods emphasize the design of iterative networks while insufficiently addressing the use of image priors. This study introduces a novel deep unfolding architecture for image CS, called GPD-CS, which integrates generative priors (GPs) and diffusion models (DMs). The proposed method involves a two-stage training process. In the first stage, a modified variational autoencoder is designed to capture compact prior representations of natural images, and a dynamic proximal mapping module is incorporated within the DUN to integrate prior information. During the second stage, the GP learning model's parameters are fixed, and a lightweight DM is employed to infer compact representations, which are then embedded into the iterative reconstruction process. Experiments validate the superior reconstruction performance of the proposed approach. The source code and pre-trained models are publicly accessible at:https://github.com/nkbourne/GPD-CS.